Xavier SoM module orientation in space

or any other effect on the thermal design of the SoM

in example we’re simply interested if it’s ok to install SoM in other durection that when Samtec connector is facing down? ie. Samtec facing up or to the side?

many thanks
Daniel

…correction, not Samtec but Molex connector :-)

Heat flow through metal is not impacted, much, by gravitational field, and none of the cooling depends on internal airflow, so orientation should have little impact - unless you have designed a passive heatsink that requires convection currents to pull air through it. The module itself won’t care about orientation.

Please see page 3 (pdf page 8) of the user’s guide, which states (as of version 2.0):

Have you looked at the Thermal Design Guidelines document? It’s available for download and goes through a lot of detail on how to design a thermal interface. If you look closely at the assembly (the 3D model is also available for download) and the TDG, you will realize there is no heat-pipe.

FWIW, heat-pipes are also fairly immune to gravity - it’s not about ‘warm air rising’, it’s about water evaporating into a space from the hot component, and condensing from that volume where the fins are cooled - that condensed water is wicked back to the hot surface. There is no gravitation required for this, it will work in 0G.

If you are wondering about a liquid/vapor phase change heat sink, then you can stop worrying…it’s just a regular heat sink on the dev kit, and you would have to add your own heat sink solution for a module sold separately. In the case of moisture condensation (if extreme cold or operating below the dew point), then I’d consider conformal coating anything you worry about.

Thx a lot . The page 9 of Development kit is sufficient evidence that the orientation is not important (I did not checked the whole kit manual before as we’re designing only with SoM module).

For the other comments: I humbly disagree - there is heatpipe inside the module (take a screwdriver and have a look if you have doubts) and also some of the liquid/vapor phase change heat pipes has dependency on vector of gravity (and it’s nowhere documented what type is used here)

points out that there is only a thermal transfer plate (TTP) in the module - no heatpipes. The dev kit uses a different thermal solution. All this information about the module is available and it is clearly described in the thermal design guidelines document. You can add your own heat pipe solution.

I’ve tested heat pipes in various orientations and the difference in cooling was not measurable. This may not be the case for all heat pipes, but keep in mind that it’s really hard to get convection currents caused by temperature differentials inside a 1mm tube that’s several cm long. Circulation in there is really driven by the phase change cycle.

What is then this if not heatpipe? :-)

Lol. Yes, I mistook the statement that the cooling interface for the module was different than the dev kit to indicate that there is no heat pipe. But looking again at the TDG, there’s an unlabeled part in the assembly which is almost certainly a heat pipe.

Regardless, with capillary action drawing the fluid along the inner surface (through scintered metal) of the tubes, and phase change driving the circulation, physical orientation of the module would be unimportant. Orientation of the dev kit matters for air-flow.

A heat pipe is not necessarily containing a liquid for phase change. However, ignoring for the moment whether or not this particular pipe has a coolant in it, consider the following…

Phase change liquids tend to require convection to send the gas phase upward, and the liquid phase downwards (and is the basis as to why the original question matters). Check the default orientation of the dev kit sitting on a desktop:
https://developer.nvidia.com/sites/default/files/akamai/embedded/images/jetsonXavier/Xavier-White_Cropped.jpg

This image shows the copper tubes are all parallel when placed on a flat table, and do not have the possibility of vapor and liquid being useful even if the pipes contain this medium.

Now if we were to orient the module in any other direction, then the liquid portion would never cover the contact point with the module. Only gas phase would exist there, and the liquid would be elsewhere (which would be the reverse of useful).

As thin as that pipe is I believe this was only used as an ordinary copper heat sink since it is easy to manufacture using tube. The ends of the tube don’t even appear to be hermetically sealed. This solution should not care about orientation, but if it does have liquid content for phase change, then there was a mistake in the design. The design is incompatible with liquid phase change.

Hi daniel,

There is water in the heat pipe, which will evaporate and move to the cooler extremities of the pipe. The heat pipe acts as a sprint to minimize the TIM bond line between the die and the copper plate which also allows a good thermal connection of the Tegra chip to the TTP. There are also direct connections between the memory and power supply inductors to improve thermal dissipation.

With regard to the orientation there are two heat pipes both in a U shape so I don’t believe that there is much change in thermal performance driven by the orientation of the device, but will confirm (update: confirmed orientation should not be an issue).

The biggest impact on thermal performance is the actual load running on the Tegra chip.

Is it correct that the water will always be in liquid phase and that there is no dependence upon vapor/liquid convection?

It’s liquid until it evaporates, that’s the whole idea of heat pipes. It’s not based on convection: fluid evaporates at the hot end and condenses at the cool end. That cycle drives the movement. The liquid is wicked by capillary action back to the hot surface.

The wikipedia article on heat pipes is very informative - including that they’re used in space applications where there is no concept of up or down.

Hi

Can I get back to this topic of heat-pipe?

We did some testing in climate chamber and we noticed quite high non-linearity with the temperature (we get much higher delta T at below zero temperatures than plus temperatures)
=> I’m wondering if you can share more information about the technology of the heatpipe used. Namely what is the medium in the heatpipe?
You wrote above “There is water in the heat pipe”, is it really water or some other medium (ie. methanol based)? According to info I found the water would not work under some 25C…

Many thanks
Daniel

Hi Daniel,

Please open a new topic for your issue. Thanks

Ok => created new topic https://devtalk.nvidia.com/default/topic/1071022/jetson-agx-xavier/heatpipe-technology/

@kayccc or @Trumany Is it possible to get the manufacturer/part number for the heat pipe assembly pictured above?